Like traditional program code, software models are not resistant to change, but evolve over time by undergoing continuous extensions, corrections, and modifications. In model-driven engineering (MDE), evolution is multidimensional leading to the model management tasks of synchronization, versioning, and co-evolution. Whereas each of these tasks has recently received increased research interest, a systematic comparison and evaluation of the different approaches is missing. Within the FAME project, we aim at establishing a uniform framework characterizing changes and their impacts. The resulting findings will provide the basis for a suite of efficient techniques for avoiding unexpected side-effects of evolution. We will use different, well-explored formalisms with powerful inference engines exploiting concise semantic definitions of the modeling languages. By this, FAME will contribute to reliable change propagation indispensable for automatic quality assurance in MDE.